Trino at scale: health checking, monitoring, and observability

[dprophet]

As we scale Trino larger and larger, multiple clusters, we need near real-time health checking, monitoring and observability of Trino. Not just Trino but even the health/connectivity of the underlying catalog system.

I am wondering if other Trino users are running into these needs? Here is a starting list of our requirements. I am wondering if others have addressed these issue.

• Query stats
• User stats CPU/Memory
• Catalog health check
• Kubernetes health
• SQL errors
• Trino crash stacks

We know of these tools in the industry. Anyone using them?

• https://icinga.com/
• https://www.zabbix.com/
• https://www.nagios.com/
• https://grafana.com/
• https://www.nagios.com/
• https://prometheus.io/